85

Am. J. Trop. Med. Hyg., 64(1, 2)S, 2001, pp. 85–96Copyright� 2001 by The American Society of Tropical Medicine and Hygiene

THE ECONOMIC BURDEN OF MALARIA

JOHN LUKE GALLUP AND JEFFREY D. SACHSCenter for International Development, Harvard University, Cambridge, Massachusetts

Abstract. Malaria and poverty are intimately connected. Controlling for factors such as tropical location, colonialhistory, and geographical isolation, countries with intensive malaria had income levels in 1995 of only 33% that ofcountries without malaria, whether or not the countries were in Africa. The high levels of malaria in poor countriesare not mainly a consequence of poverty. Malaria is geographically specific. The ecological conditions that supportthe more efficient malaria mosquito vectors primarily determine the distribution and intensity of the disease. Intensiveefforts to eliminate malaria in the most severely affected tropical countries have been largely ineffective. Countriesthat have eliminated malaria in the past half century have all been either subtropical or islands. These countries’economic growth in the 5 years after eliminating malaria has usually been substantially higher than growth in theneighboring countries. Cross-country regressions for the 1965–1990 period confirm the relationship between malariaand economic growth. Taking into account initial poverty, economic policy, tropical location, and life expectancy,among other factors, countries with intensive malaria grew 1.3% less per person per year, and a 10% reduction inmalaria was associated with 0.3% higher growth. Controlling for many other tropical diseases does not change thecorrelation of malaria with economic growth, and these diseases are not themselves significantly negatively correlatedwith economic growth. A second independent measure of malaria has a slightly higher correlation with economicgrowth in the 1980–1996 period. We speculate about the mechanisms that could cause malaria to have such a largeimpact on the economy, such as foreign investment and economic networks within the country.

POVERTY AND MALARIA

Malaria and poverty are intimately connected. As T. H.Weller, a Nobel laureate in medicine, noted, ‘‘It has longbeen recognized that a malarious community is an impov-erished community.’’1 Weller could have said the same formalarious countries. Malaria is most intractable for countriesin the poorest continent, Africa. The only parts of Africafree of malaria are the northern and southern extremes,which have the richest countries on the continent. India, thecountry with the greatest number of poor people in theworld, has a serious malaria problem. Haiti has the worstmalaria in the Western Hemisphere, and it is the poorestcountry in the hemisphere.

Malaria risk has always been geographically specific, asshown in Figure 1. Intensive malaria is confined to the trop-ical and subtropical zone. Poverty is also geographically spe-cific. As shown in Figure 2, poor countries predominate inthe same regions as malaria. Almost all of the rich countriesare outside the bounds of intensive malaria.

A basic problem when studying the macroeconomic im-pact of malaria is the lack of high-quality data on malariaincidence or prevalence in the most severely affected coun-tries. This study uses an index of malaria prevalence derivedfrom historical maps of the geographical extent of high ma-laria risk shown in Figure 1 (digitized from maps by Pam-pana and Russell2 and the World Health Organization3,4

[WHO]). Combined with detailed data on the world popu-lation distribution, one can estimate the fraction of the pop-ulation in high malaria risk areas in each country.5 Becausemost malaria mortality and severe morbidity is due to oneof the 4 malaria species, the malignantPlasmodium falci-parum, the index of malaria intensity used in this article isthe fraction of the population at risk of malaria multipliedby the fraction of cases of malaria that are falciparum ma-laria (from WHO data6). (A second index of malaria derivedfrom completely different data is described and used below.)For the comparative statistics in this section, severe malariais defined as having a malaria index of� 0.5.

After the first draft of this article was completed, McCar-thy and others (unpublished data) have estimated the impactof malaria on economic growth by use of recently releasedestimates of malaria morbidity from the WHO.7 By use of asimilar methodology to ours, but by use of a different datasource and a different time period, they found somewhatsmaller estimated effects of malaria than we find here. Thesmaller estimates could be due to measurement error in theWHO data because the data ‘‘permit only very limited com-parison between countries or even between various periodsfor the same country.’’7 Because the national reporting sys-tems are systematically different between countries withhigh or low levels of malaria, this study does not use theWHO data on cases of malaria but instead uses the malariaindex derived from malaria maps and falciparum prevalencedata.

As an example, take the 150 countries with populations� 1 million in 1995, which account for� 99% of the world’spopulation. Forty-four of the 150 countries, or 29%, haveintensive malaria. Thirty-five of these 44 countries are inAfrica. The average purchasing-power parity gross domesticproduct (GDP) per capita in 1995 for the malarial countrieswas $1,526, compared with an average income of $8,268 inthe countries without severe malaria,� 5 times higher. (The1995 purchasing-power parity GDP data are from the WorldBank, supplemented by U.S. Central Intelligence Agency es-timates for countries not reported in World Bank data.8–10)Ranking the 150 countries by income per capita, all but 3of the 44 countries with severe malaria are in the bottomhalf of the ranking. The exceptions are Oman and Gabon,ranked 34th and 41st, which owe their wealth to oil, and thePhilippines, which is barely in the top half, with a rank of74 out of 150. Of the 119 poorest countries, all but 12 havesome incidence of malaria or are recovering from socialism.The richest 31 countries are free of malaria as measured byour index.

Not only are malarial countries poor, but economic growthin malarial countries over the past quarter century has been

86 GALLUP AND SACHS

FIGURE 1. Malaria risk, 1946, 1965, 1944.

FIGURE 2. Gross domestic product (GDP) per capita, 1995.

87ECONOMIC BURDEN OF MALARIA

TABLE 1Level of gross domestic product (GDP) per capita

Regression model

1 2 3 4 5 6

Variable

Log GDP per capita†

1995 1950 1995 1995 1995 1995 (non-Africa)

Population within 100 km of coast (%)

Log distance to major markets

Log hydrocarbons per person

Tropical land area (%)

1.26(6.31)**

�0.35(3.79)**0.01

(2.28)*�0.68(3.97)**

0.80(5.19)**

�0.12(1.37)0.01

(2.56)*�0.14(0.89)

0.57(2.74)**

�0.33(4.03)**0.01

(1.86)�0.23(1.01)

0.65(3.40)**

0.01(2.13)*

�0.59(3.04)**

0.33(2.23)*

0.00(1.36)

�0.09(0.59)

0.40(2.76)**

0.00(1.27)

�0.10(0.83)

Falciparum malaria index

Socialist

Colony

Trade openness (0–1)

�1.17(6.28)**

�1.22(5.67)**

�1.16(4.73)**

�0.80(5.20)**

�0.14(2.18)*

�1.16(6.41)**

�0.10(0.66)

�0.05(0.89)0.50

(2.99)**

�1.10(4.34)**

�0.05(0.30)

�0.12(2.24)*0.43

(2.98)**Quality of public institutions (0–10)

Constant

ObservationsR2

10.50(14.10)**

1490.47

8.54(13.54)**127

0.59

10.91(17.36)**127

0.62

8.75(46.40)**149

0.62

0.22(6.85)**7.15

(29.27)**97

0.88

0.237.82)**7.15

(32.30)**66

0.88

† Robust t-statistics are in parentheses.* Significant at 5% level.** Significant at 1% level.

dismal. Growth of income per capita 1965–1990 for coun-tries with severe malaria has been 0.4% per year, whereasaverage growth for other countries has been 2.3%, � 5 timeshigher. (The data for GDP growth 1965–1990 for 95 coun-tries are from the Penn World Tables.11) More than a thirdof the countries with severe malaria (11 out of 29) had neg-ative growth 1965–1990.

The question is whether these dramatic correlations meanthat malaria causes poverty and low growth. We will addressthis question in 3 ways. First, we consider the correlation ofmalaria with income levels after controlling for other factorsthat are likely to affect the world distribution of income,such as geography, history, and policy. Second, we discussthe determinants of malaria risk. Unlike other important dis-eases in poor countries caused by deficient living conditions,such as diarrhea, tuberculosis, and schistosomiasis, malariais not primarily a consequence of poverty; its extent andseverity are largely determined by climate and ecology.Third, we explore the impact of malaria on subsequent eco-nomic growth. This provides the most direct evidence of thecontinuing importance of malaria as a cause of poverty.

MALARIA AND INCOME LEVELS

The coincidence of severe malaria and low incomes couldbe due to many factors besides malaria itself. It could be ageneral effect of the tropics caused by poor soils, low agri-cultural productivity, or tropical diseases other than malaria.It may capture geographical trade barriers; many malarialcountries are landlocked and far from the major centers ofworld trade. It could be an accident of history. Many ma-larial countries were colonies until recently, and the terriblemisfortunes of colonization may linger, keeping incomes

low. Malaria could simply be a proxy for Africa, which maybe poor for other reasons, such as weak institutions, pooreconomic policies, or ethnic conflict.

There are strong geographical patterns to income levelsaround the world. (See Gallup and others12 for a wider in-vestigation of the role of geography in economic develop-ment and for a more detailed explanation of the variablesused in this section.) As shown in Regression 1 of Table 1,just 4 geographical variables account for almost half the var-iation in the log of GDP per-capita income levels in 1995.A country’s accessibility to the coast, measured by the shareof the population within 100 km of the coast, is an importantindicator of success in foreign trade and integration into theglobal economy and hence is related to high income levels.Another measure of accessibility, the minimum distance tothe core world markets (New York, Rotterdam, and Tokyo),is inversely related to higher incomes. Third, resource de-posits, proxied by the log of hydrocarbon reserves per per-son, are higher in wealthier countries though the effect isvery small. Last, the tropics, measured by the percentage ofa country’s land area in the geographical tropics, is muchpoorer than the rest of the world. The ‘‘ penalty’’ for beingtropical was �0.68, signifying that tropical areas had only51% (� exp(�0.68)) of the per-capita income of nontropicalareas, controlling for other factors.

The next 3 regressions in Table 1 add the malaria indexto the geographical correlates of income per capita in 2 dif-ferent years, 1950 and 1995. (The data for purchasing-powerparity GDP per capita in 1950 in Table 1 are from Maddi-son.13 The purchasing-power parity GDP per-capita data for1995 are those used above.) The malaria index has a strongnegative association with income levels after controlling for

88 GALLUP AND SACHS

TABLE 2Level and changes in malarial prevalence between 1965 and 1994

by climate zone*

Predominant climateMalaria index,1965 (0–100)

Average change,1965–1994

Temperate (n � 57)Desert (n � 23)Subtropical (n � 42)Tropical (n � 21)

0.227.861.764.9

�0.2�8.8�5.0

0.5

* Countries are classified by their predominant ecozone from the following groupings:temperate (temperate, boreal, and polar ecozones), desert (tropical and subtropical deserts),subtropical (nondesert subtropical), and tropical (nondesert tropical).12 The index and av-erage reduction are unweighted averages over countries.

the other 4 geographical factors. Malaria’s coefficient in-creases slightly from 1950 to 1995, suggesting that if any-thing, malaria has become more important for explainingincome levels over time. The association of income withmalaria dominates the association with the tropics, whichloses statistical significance in the regressions. The coeffi-cient on malaria of �1.22 in 1995 implies that malarialcountries have a per-capita income only 30% as high as non-malarial countries.

Regression 4 includes indicators for former colonies andfor socialist countries in the post-World War II era. Thesenew explanatory variables are strongly associated with lowerincome levels, but taking them into account does not sub-stantially alter the correlation of malaria with low incomes.Regression 5 adds a measure of economic policy, tradeopenness in the 1965–1990 period, and an index of the qual-ity of government institutions. Malaria’s association is stillunaffected, but the socialist and colony variables lose theirsignificance. If malaria is excluded from this regression, in-come levels are significantly lower in countries that havebeen colonized, which suggests that the economic weaknessof countries with malaria might have been a factor in theircolonial subjugation.

The final regression in Table 1 excludes the sub-SaharanAfrican countries. Malaria has just as strong an associationwith poverty outside of Africa as for the whole world. Ma-laria is clearly distinguishable from other problems faced byAfrica.

Geography, history, and policy all have clear correlationswith income levels, but taking them into account does notalter the pattern of lower incomes in malarial countries. Theassociation of malaria with poverty seems to be more thanjust a mask for other plausible causes of low income.

CAUSE OR EFFECT?

Malaria is prevalent in the poorest countries. Could thisbe a consequence, rather than a cause, of poverty? Manyother serious diseases predominantly found in poor countriesclearly are a direct consequence of poverty, caused by in-adequate sewage treatment, unsafe drinking water, poor hy-giene, or substandard housing. Malaria, though, does not fol-low this pattern; its severity, and the difficulty in controllingit, are determined mainly by climate and ecology. Personalbehavior, such as use of screens and bed nets, and the gen-eral level of development, especially urbanization, also af-fect malaria prevalence, but they are not the main determi-nants.

Certain countries with high incomes still face serious ma-laria problems because of their geographical location. Oman,with an income per capita of almost $10,000, has severemalaria throughout the country except in remote areas ofhigh altitude and desert. United Arab Emirates, next doorwith one of the highest income levels in the world, has alsobeen unable to eliminate malaria.

Successful elimination of malaria through vector controlrequires a well-run organization and financial resources. Thedetermining factor in where malaria has been eliminated inthe postwar era has not been institutional or financial, how-ever. It has been the susceptibility of malaria and the vectorto control. Figure 1 shows that since 1946, malaria has only

been eliminated in nontropical regions and certain islandswhere it foothold is much weaker. Coluzzi14 writes, ‘‘ Aboveall, it should be stressed that malaria eradication [in temper-ate areas in the late 1940s and early 1950s] was [only]achieved within more or less marginal ecoepidemiologicalzones, particularly for P. falciparum’’ because of seasonalityof malaria transmission, low nighttime outdoor temperatures,and less efficient malaria vectors in temperate regions.

The large differences in the difficulty of controlling ma-laria in various climatic zones is supported by informationprovided in Table 2. Those regions with the worst malariain 1965 had the least reduction in malaria in the next 3decades. Countries with a predominantly humid tropical cli-mate actually saw a small increase in the malaria index. Al-though the absolute reduction in the malaria index in tem-perate countries was lower than in other climatic zones, thatis because the malaria level is bounded by zero; malaria asmeasured by the malaria index was completely eliminatedin temperate countries.

Some of the most effective control efforts historically inthe worst affected areas have used few material resourcesother than labor, so they are not constrained by poverty perse. The elimination of breeding sites for malarial mosquitoesin parts of Panama by Gorgas at the time the canal was built,the control of the outbreak of Anopheles gambiae mosqui-toes in northeastern Brazil in the 1930s, and the malaria-freeenclaves around some African mines show what is possiblewith a combination of complete monitoring of all open watersources inside and outside households, drainage of wetlands,and a military precision in all operations.15–17 Unfortunately,such control efforts have never been sustained in more thansmall areas or for more than short periods of time.

The major efforts devoted to malaria control in the build-ing of the Panama canal and at African mines demonstratesthe economic impact of malaria on workers. Malaria mor-tality was a major factor in the French failure to completethe canal (at least 20,000 people were lost in 9 years), andthe American efforts were not effective until malaria andyellow fever were brought under control.15 Some tropicalAfrican mines created a cordon sanitaire around their op-erations where African workers could not regularly leave orenter. The large investments in monitoring, drainage, andhousing could only have been justified by higher workerproductivity in the malaria-free mines.

In addition to differences in malaria intensity due to cli-mate, the world distribution of Anopheles mosquitoes—themalaria vector—have a major impact on malaria prevalenceand severity. Vectorial capacity is a measure of the efficiency

89ECONOMIC BURDEN OF MALARIA

with which mosquitoes carry malaria from one human toanother, an estimate of the number of secondary cases ofmalaria generated by one primary case. The vectorial capac-ity of different species of Anopheles varies by orders of mag-nitude. By far the most efficient vector, Anopheles gambiae,is exclusively found in sub-Saharan Africa.

Vectorial capacity has a major impact on the feasibility ofcontrolling or eradicating malaria in a region. Consequently,malaria eradication through vector control has been ordersof magnitude more difficult in sub-Saharan Africa. Accord-ing to a recent expert committee report, ‘‘ The epidemiologyof malaria is driven by the dynamics of the mosquito vectors.Thus, 90% of the world’s malaria is in Africa because it ishome to the three most effective vectors.’’ 18 Not only do themosquito species determine the intensity of transmission, butthey also affect the mix of malaria between the malignantP. falciparum and the less severe Plasmodium vivax, Plas-modium malariae, and Plasmodium ovale. Africa is also theonly major region of the world where falciparum malariapredominates.

Malaria control in sub-Saharan Africa has been a non-starter. There has been no successful malaria control of largeregions outside of the temperate southern tip, the controlledenvironment of some mining camps, and a few islands. Inresponse to the failure of WHO vector control projects inCameroon, Nigeria, and elsewhere in Africa in the 1960s,the WHO sponsored an intensive malaria control and re-search project in the district of Garki, Nigeria.19 No resourc-es, manpower, or institutional support were spared. Over thecourse of 7 years, WHO and the Nigerian government spentmore than $6 million to try to eliminate malaria in 164 vil-lages and compare the changes to control villages. Insecti-cide spraying of every hut at least every 10 weeks duringthe course of the study had an average coverage of 99%. Athird of the villages were also given mass drug administra-tion as a prophylaxis against malaria.

The intensity of malaria transmission in Garki was ‘‘ veryhigh indeed.’’ 19 During the wet season, a person in this dis-trict would be bitten on average 174 times per night by theAnopheles gambiae s.l. malaria vector and 94 times per nightby the Anopheles funestus vector. Such high biting rate es-timates are not unusual. Robert and others20 estimate that aperson in the Kou Valley in Burkina Faso sleeping withoutmosquito protection—as most do—receives 158 bites byAnopheles gambiae per night, with total mosquito bites of35,000 per year. The vectorial capacity, or the transmissionrate of malaria between people through the vectors, reached2,000 times the critical value required to maintain endemicmalaria, with a range of 18–145 malaria-transmitting bitesper person per year in the 8 villages studied.20 In lay terms,everyone was constantly reinfected with malaria.

The vector control efforts reduced the human-biting rateof mosquitoes in the Garki villages by 90% from their pre-study level, but despite this huge reduction in mosquito den-sity, there was no significant change in the parasite rateamong the villagers. The control efforts were defeated bythe vectorial capacity of the mosquitoes, which vastly exceedwhat was required to maintain transmission of malaria. Theconclusions of the study show that the failure to control ma-laria in similar environments was not the consequence ofpoverty or lack of institutional capacity. According to a con-

ference paper summarizing the Garki study: ‘‘ The malariacontrol measures employed in the Garki Project failed tohave a significant overall impact on malaria transmission,suggesting that these measures are unlikely to be of long-term use in the African dry savannah belt.’’ This failureoccurred despite the fact that ‘‘ at all times during this study,it was known that the strategies employed were much toodetailed and expensive for long-term use in the study area’’(Loutan and others, unpublished data).

At least 2 biological factors explain the exceptional se-verity of malaria in Africa. The most efficient mosquito vec-tor and the most serious malaria strain both most likely camefrom Africa. The vector Anopheles gambiae s.s. coevolvedwith humans in the Afrotropical rain forest. The develop-ment of African agriculture in forest clearings resulted in thevector’s most important characteristic for malaria transmis-sion: it almost exclusively bites humans.14 The explosive po-tential of the Anopheles gambiae vector for transmitting ma-laria in similar climates elsewhere was shown by the acci-dental introduction of the mosquito into Brazil in the late1920s, which was luckily brought under control soon enoughto eliminate it.16 The most pathogenic human malaria spe-cies, P. falciparum, most likely originated in Africa, prob-ably in the past 5,000–10,000 years with the onset of agri-culture.14

With no proven method of controlling malaria in sub-Sa-haran Africa and other areas of intense transmission, it isdifficult to argue that poverty effectively causes malaria ordetermines the success of control efforts. A recent U.S. Na-tional Institutes of Health report18 notes the intractable natureof malaria Africa: ‘‘ The availability of anti-malaria mea-sures, when correctly integrated and applied without finan-cial constraints, can probably cope successfully with the ma-laria problem everywhere in the Tropics except in the Af-rotropical region’’ (emphasis added).

A different sort of evidence that malaria is a cause ofpoverty comes from evolution. In areas with the most severemalaria today, sub-Saharan Africa and parts of the MiddleEast and India, many ethnic groups have developed a partialgenetic defense against the ravages of malaria: sickle-celltrait. In some parts of Africa, this red blood cell abnormalityis carried by 25–30% of the population.21 The value of sicklecell’s protection against malaria must be great because itcomes at a high cost: all children in developing countrieswho inherit the trait from both mother and father die beforethey reach childbearing age. The Garki project confirmedthis cruel equilibrium.19 Sickle-cell trait in Garki adults wasmuch higher than in children because of selective survival.The burden of malaria on human well-being must have beenhigh indeed for such a mutation to be beneficial. Sickle-celltrait also shows the role of climate in determining the rela-tive burden of disease in different regions of the world: ‘‘ Thedistribution of sickle cell trait in tropical Africa correspondsalmost exactly to the areas of tropical rain forest.’’ 22

Milder congenital blood diseases, such as thalassemia inparts of southern Europe and Asia, confer some protectionagainst malaria in regions where malaria is correspondinglyless severe. These blood diseases highlight the importanceof the burden of falciparum malaria relative to other formsbecause they protect primarily against falciparum infec-tions.23 (Many ethnic groups in Africa also have complete

90 GALLUP AND SACHS

TABLE 3Gross domestic product (GDP) per-capita growth before and after

malaria eradication in southern European countries (late 1940s)

Country

GDP growth

1913–1938 1950–1955

Difference withwestern Europe

1913–1938 1950–1955

GreeceItalySpainWestern Europe

2.11.0

�0.40.9

3.65.36.22.3

1.10.1

�1.40.0

1.33.04.00.0

TABLE 4Gross domestic product per-capita growth before and after malaria

eradication in Portugal (1958)

Country 1953–1958 1958–1963 Change

PortugalWestern EuropeDifference

3.01.9

�1.1

5.33.8

�1.5

�2.3�1.9�0.4

TABLE 5Gross domestic per-capita growth before and after malaria eradica-

tion in Taiwan (1961)

Country 1956–1961 1961–1966 Change

TaiwanEast AsiaDifference

2.83.4

�0.6

5.85.5

�0.3

�3.0�2.1�0.9

protection from P. vivax malaria due to a blood characteristiccalled the Duffy factor, which makes vivax malaria rare inAfrica. Although this suggests that vivax malaria is also bur-densome, it does not demonstrate that the human burden islarge because the Duffy factor causes no mortality in peoplewho carry it. Africans could easily maintain the Duffy factorin the face of evolutionary selection even with a low diseaseburden from vivax malaria.)

The geographical specificity of malaria, the wide biolog-ical variation in the capacity of mosquito vectors, the in-ability to control malaria in Africa under experimental con-ditions, and the persistence of fatal blood diseases as a de-fense all point to a causation from malaria to poverty, notvice versa. Large-scale vector control projects require re-sources, but if they were clearly feasible, the resourceswould probably be forthcoming from the international com-munity. Much of the effective malaria control (in subtropicalareas) has in fact come from low-technology drainage andlarviciding, which could be carried out independently by apoor tropical country if the technique offered a viable pros-pect of malaria control. Kriton and Spielman24 describe themajor role of these simple technologies in many of the suc-cessful eradication efforts.

ANECDOTES FROM COUNTRIES THAT HAVE ELIMINATED MALARIA

A small number of the countries that had severe malariain the Twentieth Century eliminated the disease. Many otherchanges were simultaneously occurring in the economies ofthese countries before and after eradication, but in almost allcases for which we have data, the countries experienced anacceleration of growth immediately following eradication ofmalaria, and faster growth than neighboring countries.

Malaria eradication in southern Europe has been a clearsuccess story in the fight against malaria. Major control ef-forts in Greece, Italy, and Spain were started in the 1930sand completed in the late 1940s. Greece up to that time hadbeen the most malarial country in Europe; in peak years, aquarter of the total population was infected.25 Jones26 arguesthat the spread of falciparum malaria through most of Greecein the first millennium was the main factor in the decline ofancient Greek civilization. Greece was the site of major ma-laria epidemics in the 19th and early 20th century, and thefamed plain of Marathon became virtually uninhabited dueto malaria, despite fertile soil. The use of DDT (dichloro-diphenyltrichloroethane) starting in 1946 had spectacular re-sults (which in turn had a major influence on the subsequentWHO world eradication campaign), with malaria fallingfrom 1–2 million cases per year in the early 1930s to only

5,000 cases in 1951.27 Although complete eradication wouldtake another 20 years, partly because of vector resistance toDDT, from an economic point of view, malaria was undercontrol.

The long-standing problem of malaria in Italy contributedto the major role of Italians in early malaria research. Justbefore the control campaign, Italy had � 300,000 cases ofmalaria per year, with �20,000 deaths.28 The Pontine Marsh-es south of Rome were rendered uninhabitable by the dis-ease. Plasmodium falciparum was eliminated by the end ofthe 1940s, with P. vivax and P. malariae disappearing moreslowly. Spain reported 400,000 cases of malaria with 1,700deaths in 1943, but it had effectively controlled the diseaseby the end of the 1940s.28

The period immediately before effective control of malar-ia was wartime and the postwar reconstruction. Because ofthe anomalies of the period and the lack of data, we comparegrowth in the postcontrol years of 1950–1955 to growth inthe period 1913–1938 in Table 3. (GDP data for the 1913–1938 period are from Maddison.13 All other country GDPdata in this section are from Summers and Heston.11) In all3 countries, economic growth in the postcontrol period wasmuch higher than in the prewar period and higher thangrowth in rest of western Europe in 1950–1955. In the pre-war period, Greece and Italy also grew somewhat faster thanwestern Europe, but the increment in growth over the Eu-ropean average was also higher in the postcontrol periodthan the prewar period.

Portugal was another southern European country with se-vere malaria (over 100,000 cases per year in the 1940s) thatcontrolled malaria later than Greece, Italy, and Spain.28 Asshown in Table 4, growth accelerated after eradication in1958 compared with the period before eradication, and onceagain, the increment of growth over the average in the restof western Europe increased after eradication.

There are, unfortunately, few success stories for malariaeradication in developing countries, but the islands of Tai-wan and Jamaica are among the few. Tables 5 and 6 showthat growth accelerated in the 2 countries after eradication,in 1961 for Taiwan and 1958 in Jamaica. In both cases,growth also increased by more than growth in their respec-tive regions.

91ECONOMIC BURDEN OF MALARIA

TABLE 6Gross domestic product per-capita growth before and after malaria

eradication in Jamaica (1961)

Country 1956–1961 1961–1966 Change

JamaicaCentral America and CaribbeanDifference

3.42.6

�0.8

4.13.1

�1.0

�0.7�0.5�0.2

The South of the United States was still malarious beforeWorld War II; 135,000 cases of malaria with 4,000 deathswere reported in 1935.28 After large-scale drainage projectsby the Works Progress Administration (WPA) in the 1930swere followed by insecticide spraying after the war, malariawas brought under control by the end of the 1940s. In thedecade of the 1950s, the South had its most dramatic catch-up with the rest of the country, going from 60% of the in-come per capita of the rest of the United States in 1950 to68% in 1960 (calculated from Barro and Sala-i-Martin29).

An exception to prove the rule is Mauritius. A small islandoff the coast of East Africa, Mauritius was first exposed tomalaria in 1865 with catastrophic results. In a single year,1867, between an eighth and a quarter of the total populationdied in the malaria epidemic.30,31 Malaria was finally elimi-nated in 1963. Economic growth in a small, closed, sugar-producing economy continued to be negative until 1973,when Mauritius opened its economy, built export processingzones, and took off economically. Countries do not becomeprosperous by controlling malaria alone, but the dramaticsuccess of Mauritius in become a manufacturing exportersince 1973 was certainly made easier by eliminating malaria.

Malaria control within regions of some other countries hashad dramatic impacts on agricultural output and settlementpatterns: ‘‘ Until malaria was wiped out [in Corsica], no onefarmed [on the eastern plain]. Today this plain accounts for60 percent of Corsica’s agricultural production.’’ 25 Thesouthern plains of Nepal, the Terai, were virtually uninhab-ited until the early 1950s because of malaria. It is now therichest and most agriculturally productive part of the coun-try.25

These country examples of growth after the control ofmalaria are merely suggestive. In almost every country weexamined, economic growth was higher immediately afterthe eradication of malaria, but there were surely many otherfactors that influenced the economy at the same time. Inseveral of the countries (Greece, Spain, and Jamaica), therapid development of the tourism industry was only possiblebecause of malaria eradication. Few tourists thought of bask-ing on shores of the Aegean when Greece was the mostmalarial country in Europe.

MALARIA AND ECONOMIC GROWTH

We have shown that most malarial countries are poor, andcertain countries that managed to completely eliminate ma-laria in recent times have had more rapid economic growththan their neighbors. But can we find any general, statisti-cally convincing evidence that initial malaria prevalence andreductions in malaria affect economic growth? Would a re-duction in malaria significantly improve the economic pros-pects of poor countries?

The most direct way to assess the causal effect of malariaon country economic performance is to look at the relation-ship between economic growth, initial malaria levels, andchange in malaria over the same period. Above, we saw thatcountries with severe malaria in 1965 have had much lowereconomic growth in the subsequent 25 years, but this didnot take into account the initial poverty of countries, nor didit consider the role of human capital levels, government pol-icies, or geographical variables. After the role of human cap-ital, policy, and geography are taken into account, it is gen-erally found that poorer countries grow faster than richercountries, so if malaria were really just a proxy for poverty,one would expect malarial countries also to grow faster.32 (Infact, over the 1965–1990 period, poor countries on averagegrew slower than rich countries, but poor countries also hadlower initial human capital, followed less successful eco-nomic policies, and were disadvantaged geographically.)

Table 7 presents a cross-country empirical growth esti-mation in the style of Barro.33 Growth in GDP per capitaover the 1965–1990 period is related to initial income levels,initial human capital stock, policy variables, and geograph-ical variables. Human capital stock is measured by second-ary education and life expectancy at birth. Policy is mea-sured by trade openness over the period and an index of thequality of public institutions. The geographical variables in-clude an indicator for the geographical tropics and the frac-tion of the population within 100 km of the coast. (Gallupand others12 give a more detailed description of these vari-ables.) To these well-researched predictors of economicgrowth, we add the malaria index in 1965 in Regression 1.

The malaria index for 1965 (Figure 3) is constructed sim-ilarly to the malaria index described above. It is the productof the fraction of the population living in areas with highmalaria risk in 1965 times the fraction of malaria cases in1990 that are due to P. falciparum.3,5,6 This assumes that therelative share of P. falciparum cases did not change sub-stantially from 1965 to 1990. The change in the malaria in-dex over the 1965–1994 period was constructed with a sim-ilar malaria index for 1994 (Figure 4).4–6

Countries with severe malaria in 1965 had much lowereconomic growth, amounting to 1.3% lower growth per year,even after other factors such as initial income level, overallhealth, and tropical location are taken into account.

Reductions in malaria over the 1965–1990 period, in ad-dition to malaria levels in 1965, are associated with muchhigher economic growth, as shown in Regression 2 in Table7. This corresponds to a 0.3% rise in annual economicgrowth for a 10% reduction in the malaria index. Over the25-year period, the average reduction in the malaria indexwas 7% among countries that had malaria in 1965. By ex-trapolation far outside the observed sample variation, acountry with its whole territory affected by 100% P. falci-parum malaria is predicted to permanently raise its annualgrowth by 2.6% if it completely eliminates malaria! Unfor-tunately, no country came near to accomplishing this. Of the14 countries in the sample with a malaria index � 0.9 in1965, only one reduced it significantly: the malaria index inZimbabwe fell by one third.

Economic growth itself might be a cause of the observedmalaria reductions if greater resources were made availablefor malaria control, or if a high institutional capacity were

92 GALLUP AND SACHS

TABLE 7Growth of gross domestic product (GDP)

Regression model†

3 4

Variable1

1965–19902

1965–19901965–1990

(IV)1965–1990

(non-Africa)5

1965–19906

1980–1995

Log initial GDP per capita

Log initial secondary schooling

�2.6(8.07)**0.1

(1.04)

�2.6(7.90)**0.1

(0.90)

�2.4(7.54)**0.1

(0.60)

�2.5(6.36)**0.1

(0.62)

�2.3(8.04)**0.1

(0.77)

�3.6(7.95)**

�0.2(0.62)

Log initial life expectancy

Trade openness (0–1)

Quality of public institutions (0–10)

4.4(4.46)**1.8

(4.91)**0.4

3.1(3.41)**1.7

(4.91)**0.4

3.0(3.51)**1.6

(4.51)**0.3

3.8(2.34)**1.7

(4.14)**0.4

4.6(4.19)**1.7

(4.55)**0.3

9.6(3.44)**3.0

(5.10)**0.6

Tropical land area (%)

Population within 100 km of coast (%)

Initial falciparum malaria index

(3.29)**�0.6(1.30)0.9

(2.85)**�1.3(2.24)**

(3.79)**�0.6(1.31)0.7

(2.64)**�2.1(3.77)**

(3.32)**�1.0(2.55)**0.7

(2.41)**�1.8(3.12)**

(2.95)**�0.6(1.28)0.6

(1.66)�1.8(1.77)

(2.78)**�1.0(2.50)**0.8

(2.36)**�1.3(1.98)**

(4.03)**�0.6(1.22)0.9

(1.80)

Change of falciparum malaria index

Tropical disease, first principle component

Initial World Health Organization advisory malaria index

Constant 1.3

�2.6(4.07)**

6.1

�2.5(3.48)**

5.7

�2.2(2.24)**

3.7

0.1(1.51)

�0.9

�1.6(2.8)**

�14.8

ObservationsR2

(0.36)75

0.77

(1.68)75

0.80

(1.58)73

0.80

(0.63)60

0.76

(0.21)73

0.77

(1.42)780.71

† Robust t-statistics are in parentheses.* Significant at 5% level.** Significant at 1% level.

FIGURE 3. Malaria index, 1965.

93ECONOMIC BURDEN OF MALARIA

FIGURE 4. Malaria index, 1994.

responsible both for economic growth and successful malariacontrol. In this case, the estimates of the effect of malariareduction on economic growth would be biased. To controlfor the possible endogeneity of malaria reduction, Regres-sion 3 in Table 7 uses instrumental variables. The instru-ments are the prevalence of 53 different Anopheles mosquitovectors in each country in 1952. (The Anopheles data weredigitized from an American Geographical Society map34 andused to calculate the percentage of land area in each countryaffected by each Anopheles species.) The different Anophe-les mosquitoes vary widely in their efficacy in transmittinghuman malaria, so that the distribution of Anopheles vectorsis strongly correlated with malaria intensity and its change(the first-stage regression of the change in the malaria indexon Anopheles vectors has an R2 of 0.51). There is no reasonto think that the distribution of malaria mosquito vectors isa cause of economic growth apart from the direct influenceof malaria, making vector prevalence an ideal instrument formalaria change. After correcting for the possible endogene-ity of malaria reduction, the estimated effect on economicgrowth is essentially unchanged, so it is unlikely that thechanges in malaria prevalence are a consequence of econom-ic growth. A Hausman test finds no significant difference theordinary least-squares and instrumental variables estimates,rejecting the endogeneity of the change in malaria.

Regression 4 in Table 7 restricts the sample to non–sub-Saharan African countries. The size of the estimates for ma-laria are substantially the same. The change of malaria hasa statistically significant coefficient, but the estimate for ini-tial malaria loses statistical significance. Even without in-cluding the sub-Saharan African countries with the most se-

vere malaria, a reduction in malaria corresponds to muchhigher economic growth.

Malaria could be a proxy for a range of tropical diseasesthat are not adequately controlled for by life expectancy. Onedisease that is starting to have major economic impacts inmany of the same countries with severe malaria, acquiredimmunodeficiency syndrome (AIDS), is not relevant for thetime period under study here. By 1990, the end of the periodof economic growth studied, the burden of AIDS was stillsufficiently small to cause only minor economic impacts.Other major diseases prevalent in the tropics that may becorrelated with malaria are hookworm, onchocerciasis,schistosomiasis, filariasis, dengue fever, and trypanosomiasis(sleeping sickness). We have identified detailed maps of thegeographical extent of all these diseases except for trypano-somiasis from the 1950s, as well as data for 10 other, lessimportant tropical diseases providing measures of the extentof 20 different tropical diseases.35–37 The other diseases aredengue fever, yellow fever, helminthiases (Paragonimuswestermani, Fasciolopsis buski, Opisthorchis felineus, Di-phyllobothrium latum, and Clonorchis sinensis), and leish-maniases (oriental sore, kala azar, and American). The schis-tosomiasis data are broken down into Schistosoma haema-tobium and Schistosoma mansoni, and the filariasis data intoloa loa, Wucheria bancrofti, Wucheria malayi, Acanthochei-lonema perstans, and Mansonella ozzardi, giving a total of20 nonmalaria tropical disease variables. The land area af-fected by the disease is weighted by detailed population dis-tribution data (in 19945) to provide an estimate of the frac-tion of the population at risk of each disease. Because thedisease data precede the period of economic growth under

94 GALLUP AND SACHS

study, they show the impact of initial disease on later eco-nomic performance, thus avoiding problems of reverse cau-sation.

The large number of diseases makes it impractical to in-clude them all as independent correlates in the economicgrowth regression due the limited sample of countries. Toassess whether the other diseases were responsible for thecorrelation of initial malaria with economic growth, we in-cluded each of the 20 diseases as an additional regressorseparately to the regression specification in Regression 1 ofTable 7. The estimated impact of malaria was remarkablystable across these 20 regressions, with a point estimaterange of just �0.7 to �1.3, and statistically significant at the10% level in 17 of 20 regressions (data not shown). Rathersurprisingly, none of the other tropical diseases had a sig-nificant negative correlation with economic growth, even atthe 10% level, after controlling for malaria in these regres-sions. A second way to combine the other disease informa-tion is to estimate its principal components and include alinear combination of the other disease variables in thegrowth regression. As shown in Regression 5 of Table 7, thefirst principal component of the tropical diseases has an in-significant positive correlation with subsequent economicgrowth, and malaria has the same significantly negative cor-relation with economic growth as in Regression 1. Control-ling for a range of other tropical diseases does not substan-tially affect the correlation of initial malaria with subsequentgrowth.

The malaria index—although it is the best measure of ma-laria burden we could construct—is admittedly crude. Wehave also developed an alternative measure of malaria in-tensity, which, although also crude, is derived from com-pletely different data sources and covers a different timeperiod. The alternative malaria indicator used qualitative as-sessments of the severity of malaria from the WHO’s coun-try-specific health advice for travelers.38 The earliest descrip-tions of malaria in these advisories date from 1980, and theindex is set equal to 1 for countries in which malaria affectsthe whole country or the whole country except for majorcities, and the index is set at 0 otherwise. The WHO advi-sory malaria indicator for 1980 is correlated with economicgrowth across countries for 1980–1995 by use of WorldBank purchasing-power parity GDP per capita.8 Barro andLee39 provide data for secondary schooling of those aged 15and over in 1980, and the United Nations40 provides data forthe life expectancy at birth, supplemented with governmentyearbook estimates for Taiwan. The other covariates comefrom regressions covering the 1965–1990 period.

In Regression 6 of Table 7, the malaria indicator for 1980shows a significant negative correlation of initial malariawith subsequent growth. Countries with malaria throughoutthe country except for major cities had 1.6% lower growthin GDP per capita in the 1980–1995 period. By using ma-laria data from a completely independent source, and by as-sessing a different (though overlapping) time period fromthe other growth regressions, malaria was found to still havea large and statistically significant correlation with economicgrowth.

A recent study of the macroeconomic impact of malaria(McCarthy and others, unpublished data) uses still differentdata sources for measuring malaria, a different period of eco-

nomic growth (1983–1998), and different independent con-trol variables. The study finds a robust correlation betweenmalaria and growth by use of WHO morbidity data, but acorrelation of a smaller magnitude than we find here: justover one quarter of a percent per year of economic growthfor about a quarter of the sample. As discussed above, thesmaller correlation may be due to high measurement errorin the WHO malaria data.

The growth regression results show that countries withsevere malaria in 1965 had dramatically lower economicgrowth in the next 25 years, after controlling for other fac-tors that likely influenced growth, such as initial poverty,economic policy, initial health and education levels, andtropical location. Countries that managed to reduce malariaover the time had much higher economic growth. Theseproblems affected sub-Saharan Africa most severely becausemalaria levels are highest there, but the same relationshipwith economic growth holds in the non-African world. Useof an independent malaria measure over a different time pe-riod shows a similar correlation of malaria and economicgrowth.

COULD MALARIA HAVE SUCH A LARGE IMPACT ON ECONOMIC

GROWTH?

We have presented several kinds of evidence suggestingthat malaria has large economic effects. What are the chan-nels through which malaria could be a major drag on theeconomy?

The traditional medical view of malaria at its most severein holoendemic areas is that malaria contributes significantlyto child mortality and can cause acute disease in pregnantwomen, but it does not have large effects on the fitness ofother mature adults because of their partial immunity ac-quired through constant reinfection. McGregor41 states thisclearly: ‘‘ in adult life . . . a host-parasite balance resemblingcommensalism is achieved. Despite sustained infectiouschallenge, adults constitute an economically viable work-force capable of coping with the strenuous physical activitiesthat are required to maintain essential food supplies in sub-sistence agricultural communities.’’ Though this view maybe shared by many in the medical field, it has rarely beenthe subject of careful research. One wonders if the medicalfocus on mortality and acute disease obscures a general de-bilitation that could be caused by malaria. At least one articlereports that long-term asymptomatic malaria may be thecause of chronic pains and lassitude among Europeans inEast Africa.42

Formidable methodological and measurement problemsconfront any assessment of the impact of malaria on indi-vidual people and on households in areas with stable malar-ia. There is not even a clear method for diagnosing whichindividuals suffer from malaria. Virtually the whole popu-lation carries malaria parasites, and the density of parasitesis not a reliable measure of disease burden because of avariable immune response, which is still poorly understood.Fever symptoms are not specific to malaria. If everyone isinfected with malaria, there is no comparison group for mea-suring the impact of malaria on individuals with disease rel-ative to the healthy population.

If a clear measure of disease burden were available, one

95ECONOMIC BURDEN OF MALARIA

still faces the problem of assessing the cost of illness inextended rural households and accounting for the compen-sating behavior of other household members. It is hard toevaluate the cost of lost opportunities of household memberswho help out a person with malaria. Most attempts to di-rectly measure the lost work due to malaria (which ignorethese problems) find small impacts (Chima and Mills, andMalaney, unpublished data). Some recent studies have foundlarger measurable impacts of malaria at the household level(Cropper and others, unpublished data).43 However, the dif-ficulty in measuring who actually suffers from malaria in anenvironment where most people carry malaria parasites andthe myriad problems of measuring household response todebilitation make all the microeconomic estimates incom-plete.

Malaria has lifelong effects on cognitive development andeducation levels through the impact of chronic malaria-in-duced anemia and time lost or wasted in the classroom dueto illness. The importance of these effects is speculative,however, because their impact is virtually unstudied. Anemiadue to iron deficiency per se has been shown to affect thecognitive skills of children as well as their cognitive abilitiesin later life.44,45

It might be thought that malaria has a large impact in poorcountries because of its interaction with malnutrition. Ma-laria, along with other childhood infectious diseases, hasbeen found to exacerbate malnutrition. Surprisingly, how-ever, malnutrition probably confers some protection againstmalaria. McGregor,41 in his survey of the topic, finds that‘‘ the balance of available evidence indicates that malnutri-tion in humans is more commonly antagonistic to malaria.’’

In short, the impact of malaria on the productivity of in-dividuals in areas of stable malaria cannot be assessed withthe current state of research.

Whether or not individuals are significantly debilitated bymalaria, there are several other channels through which ma-laria could have large impacts on the economy. The first isthe impact of malaria on foreign direct investment and tour-ism. Malaria, unlike diseases resulting from poverty, doesnot discriminate between rich and poor victims. As long asmalaria protection is imperfect and cumbersome, well-to-doforeign investors and tourists may stay away from malarialcountries. A second channel through which malaria may af-fect the economy is limitation on internal movement. Thebetter educated and the ambitious who move to the largelymalaria-free cities lose their natural protection because oflack of exposure. They may be reluctant to maintain contactwith the countryside for fear of infection. Communities inunstable malarial areas may make people from stable ma-larial areas unwelcome. In general, the transmission of ideas,techniques, and development of transportation systems mayall be stunted by malaria.

CONCLUSIONS

The location and severity of malaria are mostly deter-mined by climate and ecology, not poverty per se. Areaswith severe malaria are almost all poor and continue to havelow economic growth. The geographically favored regionsthat have been able to reduce malaria have grown substan-tially faster afterward. The estimated impact of malaria on

economic growth, by use of 2 different measures of malaria,is large, but the mechanisms behind the impact are unclear.

Authors’ addresses: John Luke Gallup and Jeffrey D. Sachs, Centerfor International Development, Harvard University, 79 JFK Street,Cambridge, MA 02138.

Reprint requests: John Luke Gallup, Center for International Devel-opment, Harvard University, 79 JFK Street, Cambridge, MA 02138(e-mail: john�gallup@harvard.edu).

REFERENCES

1. Weller TH, 1958. Tropical medicine. Encyclopedia Britannica.Chicago: William Bennet, 495–497.

2. Pampana EJ, Russell PF, 1955. Malaria: A World Problem. Ge-neva: World Health Organization.

3. World Health Organization, 1966. Malaria eradication in 1965.WHO Chron 20: 286–300.

4. World Health Organization, 1997. World malaria situation in1994, part I. Wkly Epidemiol Rec 72: 269–274.

5. Tobler W, Deichmann U, Gottsegen J, Maloy K, 1995. TheGlobal Demography Project. National Center for GeographicInformation and Analysis. Technical Report TR-95-6.

6. World Health Organization, 1992. World malaria situation in1990, part II. Wkly Epidemiol Rec 67: 169–174.

7. World Health Organization, 1999. Malaria, 1982–1997. WklyEpidemiol Rec 74: 265–270.

8. World Bank, 1998. World Development Indicators 1998 CD-ROM. Washington, DC: International Bank for Reconstructionand Development.

9. U.S. Central Intelligence Agency, 1996. The World Factbook.Washington, DC: Central Intelligence Agency.

10. U.S. Central Intelligence Agency, 1997. The world factbook.Available at: http://www.odci.gov/cia/publications/factbook/index.html. Accessed August 15, 1999.

11. Summers R, Heston A, 1994. The Penn World Tables, Mark5.6. Available at: http://pwt.econ.upenn.edu/. Accessed Janu-ary 23, 2001.

12. Gallup JL, Sachs JD, Mellinger AD, 1999. Geography and eco-nomic development. Pleskovic B, Stiglitz JE, eds. World BankAnnual Conference on Development Economics 1998. Wash-ington, DC: World Bank, 127–178.

13. Maddison A, 1995. Monitoring the World Economy, 1820–1992. Paris: OECD.

14. Coluzzi M, 1999. The clay feet of the malaria giant and itsAfrican roots: hypotheses and inferences about origin, spreadand control of Plasmodium falciparum. Parassitologia 41:277–283.

15. McCullough D, 1977. The Path Between the Seas: The Creationof the Panama Canal, 1870–1914. New York: Simon andSchuster.

16. Soper FL, Wilson DB, 1943. Anopheles gambiae in Brazil, 1930to 1940. New York: Rockefeller Foundation.

17. Watson M, 1953. African Highway: The Battle for Health inCentral Africa. London: Murray.

18. National Institutes of Health, 1997. Final report, InternationalConference on Malaria in Africa: challenges and opportunitiesfor cooperation (January 6–9, 1997, Dakar, Senegal). Avail-able at: http://www.niaid.nih.gov/dmid/malaria/malafr/de-fault.htm. Accessed August 15, 1999.

19. Molineaux L, Gramiccia G, 1980. The Garki Project: Researchon the Epidemiology and Control of Malaria in the SudanSavanna of West Africa. Geneva: World Health Organization.

20. Robert V, Ouedraogo V, Carnevale P, 1991. La transmission dupaludisme humain dans un village au centre de la Riziere dela vallee du Kou, Burkina Faso. Robert V, Chippaux J-P,Diomande L, and others, eds. Le Paludisme en Afrique del’Ouest: Etudes Entomologiques et Epidemiologiques en ZoneRizicole et en Milieu Urbain. Paris: Editions de ORSTOM,5–15.

21. Weatherall DJ, 1984. Common genetic disorders in the tropics.

96 GALLUP AND SACHS

Warren KS, Mahmoud AAF, eds. Tropical and GeographicalMedicine. New York: McGraw-Hill, 88–102.

22. Carlson DG, 1984. African Fever: A Study of British Science,Technology, and Politics in West Africa, 1787–1864. Canton,MA: Science History Publications.

23. Luzzato L, 1984. Genetic factors modifying tropical disorders.Warren KS, Mahmoud AAF, eds. Tropical and GeographicalMedicine. New York: McGraw-Hill, 77–87.

24. Kriton U, Spielman A, 1989. Suppression of transmission ofmalaria through source reduction: antianopheline measuresapplied in Israel, the United States, and Italy. Rev Infect Dis11: 391–406.

25. Kamarck AM, 1976. The Tropics and Economic Development:A Provocative Inquiry into the Poverty of Nations. Washing-ton, DC: World Bank.

26. Jones WHS, 1909. Malaria and Greek History. Manchester:Manchester University Press.

27. Bruce-Chwatt LJ, de Zulueta J, 1980. The Rise and Fall of Ma-laria in Europe: A Historico-Epidemiological Study. Oxford:Oxford University Press.

28. Haworth J, 1988. The global distribution of malaria and thepresent control effort. Wernsdorfer WH, McGregor I, eds. Ma-laria: Principles and Practice of Malariology. Edinburgh:Churchill Livingstone, 1379–1419.

29. Barro RJ, Sala-i-Martin X, 1991. Convergence across states andregions. Brookings Papers Econ Activity 1: 107–182.

30. Verdrager J, Mamet R, Roche S, Klein JP, 1964. La Campagned’Eradication du Paludisme a l’Ile Maurice. Port Louis, IleMaurice: Imprimery Officiel.

31. Ross R, 1910. The Prevention of Malaria. New York: E. P.Dutton.

32. Barro RJ, Sala-i-Martin X, 1995. Economic Growth. New York:McGraw Hill.

33. Barro RJ, 1991. Economic growth in a cross-section of coun-tries. Q J Econ 106: 407–443.

34. American Geographical Society, 1951. Distribution of malariavectors. Geogr Rev 41: map.

35. American Geographical Society, 1952. Distribution of helmin-thiases. Geogr Rev 42: map.

36. American Geographical Society, 1952. Distribution of dengueand yellow fever. Geogr Rev 42: map.

37. American Geographical Society, 1954. World distribution ofleishmaniases. New York: American Geographical Society,map.

38. World Health Organization, 1981. Vaccination Certificate Re-quirements For International Travel and Health Advice toTravellers. Geneva: World Health Organization.

39. Barro RJ, Lee J-W, 1993. International comparisons of educa-tional attainment. J Monet Econ 32: 363–394.

40. United Nations, 1996. World Population Prospects 1950–2050(The 1996 Revision). New York: United Nations. Computerdiskettes.

41. McGregor IA, 1988. Malaria and nutrition. Wernsdorfer WH,McGregor IA, eds. Malaria: Principles and Practice of Ma-lariology. Edinburgh: Churchill Livingstone, 753–767.

42. Wilks NE, Turner PP, Somers K, Markandya OP, 1965. Chronicill-health from unrecognized malaria. East Afr Med J 42: 580–583.

43. Audibert M, Mathonnat J, Nzeyimana I, Henry M-C, in press.Role du Capital Humain dans l’Efficience Technique desProducteurs de Coton du Nord de la Cote-d’ Ivoire. Rev EconDev.

44. Lozoff B, Jimenez E, Wolf A, 1991. Long-term developmentaloutcome of infants with iron deficiency. N Engl J Med 325:687–695.

45. Pollit E, Hathirat P, Kotchabharkdi N, Missel L, Valyasevi A,1989. Iron deficiency and educational achievements in Thai-land. Am J Clin Nutr 50: 687–696.